Flood plain and channel dynamics of the Quinault and

Queets Rivers, Washington, USA

Jim E. O’Connora,*, Myrtle A. Jonesb, Tana L. Haluskaa

aUS Geological Survey, 10615 SE Cherry Blossom Dr., Portland, OR 97216, USAbUS Geological Survey, 1201 Pacific Ave. Suite 600, Tacoma, WA 98402, USA

Received 30 June 2001; received in revised form 6 March 2002; accepted 18 October 2002

Abstract

Comparison of historic channel migration rates, modern planform conditions, and overall sediment, wood, and flow

conditions and interactions for the Quinault River and Queets River in the western Olympic Peninsula, Washington, reveals

decadal- to century-scale interactions between gravel-bed channels and forested flood plains in temperate maritime

environments. The downstream alluvial portions of these two rivers can be divided into three reaches of different slope, flow,

sediment, and wood regimes: (i) the upper Quinault River is aggrading behind Lake Quinault, a natural lake that traps most

sediment and wood transported from the Olympic Mountain headwaters. (ii) The lower Quinault River, downstream of Lake

Quinault, transports only sediment and wood derived from reworking of flood-plain deposits and contributed from valley

margins. (iii) The Queets River has unimpeded movement of sediment and water from the mountainous headwaters to the

Pacific Ocean. Measurements of channel planform characteristics and historic migration rates and patterns show that these three

reaches have correspondingly distinct channel and flood-plain morphologies and dynamics. The aggrading and sediment-rich

upper Quinault River has the widest flood plain, widest active channel, greatest number of low-flow channels and flanking

gravel bars, and an average channel migration rate of 12.7F 3.3 m/year between 1900 and 1994. The comparatively sediment-

poor lower Quinault River has the narrowest flood plain, narrowest active channel, and lowest channel migration rate (4.0F 1.2

m/year); and most flow is through a single channel with few adjacent gravel bars. The Queets River has attributes intermediate

between the lower and upper Quinault Rivers, including an average channel migration rate of 7.5F 2.9 m/year. Flood-plain

turnover rates are similar for all three reaches, with channels eroding the flood plain at the rate of about 0.2% of the flood-plain

area per year, and with corresponding flood-plain half-lives of 300 to 500 years.

Observations from this study and previous studies on the Queets River show that channel and flood-plain dynamics and

morphology are affected by interactions between flow, sediment, and standing and entrained wood, some of which likely

involve time frames similar to 200–500-year flood-plain half-lives. On the upper Quinault River and Queets River, log jams

promote bar growth and consequent channel shifting, short-distance avulsions, and meander cutoffs, resulting in mobile and

wide active channels. On the lower Quinault River, large portions of the channel are stable and flow within vegetated flood

plains. However, locally, channel-spanning log jams have caused channel avulsions within reaches that have been subsequently

mobile for several decades. In all three reaches, log jams appear to be areas of conifer germination and growth that may later

further influence channel and flood-plain conditions on long time scales by forming flood-plain areas resistant to channel

migration and by providing key members of future log jams. Appreciation of these processes and dynamics and associated

0169-555X/02/$ - see front matter. Published by Elsevier Science B.V.

PII: S0169 -555X(02 )00324 -0

* Corresponding author. Tel.: +1-503-251-3222.

E-mail address: oconnor@usgs.gov (J.E. O’Connor).

www.elsevier.com/locate/geomorph

Geomorphology 51 (2003) 31–59

temporal and spatial scales is necessary to formulate effective long-term approaches to managing fluvial ecosystems in forested

environments.

Published by Elsevier Science B.V.

Keywords: Channel migration; River pattern; Large woody debris: log jams; Flood plains; Gravel-bed rivers; Olympic Peninsula

1. Introduction

Channel and flood-plain morphology develop from

a suite of processes involving sediment and water

movement, channel migration, flood-plain erosion and

deposition, and wood entrainment and deposition. For

alluvial rivers with forested flood plains, feedbacks

between channel instability and sediment and wood

input and deposition are important factors in control-

ling channel migration processes and resulting chan-

nel and bar morphology at the reach scale (lengths of

channel on the order of 10 to 20 channel widths; e.g.,

Keller and Swanson, 1979; Fetherston et al., 1995;

Abbe and Montgomery, 1996; Abbe, 2000). At

broader spatial and temporal scales involving several

decades and segments of rivers greater than about 20

channel widths, interactions between the channel and

forested flood plains have also been hypothesized to

control valley-scale patterns of channel instability and

flood-plain formation and morphology (e.g., Swanson

and Lienkaemper, 1982; Hickin, 1984; Sedell and

Froggatt, 1984; Gottesfeld and Gottesfeld, 1990;

Church, 1992; Piegay and Gurnell, 1997; Fetherston

et al., 1995; Abbe and Montgomery, 1996; Abbe,

2000; Gurnell et al., 2000). While there have been

numerous studies that describe such processes and

resulting channel and flood-plain morphology, espe-

cially at the reach scale, there have been few com-

parative studies of large alluvial rivers from which to

draw conclusions of how these interactions might

relate to specific conditions of flow, slope, wood

and sediment input, and how these interactions might

relate to broader-scale channel and flood-plain con-

ditions.

This paper describes evidence for rates and mech-

anisms of channel migration on the wood-rich Qui-

nault and Queets Rivers of the western Olympic

Peninsula in NW Washington state (Fig. 1). In partic-

ular, we describe how differences in channel migra-

tion rates and patterns between the two rivers likely

relate to differences in interactions among flow, flood-

plain forests, sediment flux, and other reach and

valley characteristics. Our conclusions draw primarily

from analysis of historic channel migration supported

by mostly anecdotal historic and field observations of

wood, channel, and flood-plain conditions and inter-

actions. Many of our interpretations regarding the

specific role of wood in affecting channel and flood-

plain dynamics derive in part from studies by Fether-

ston et al. (1995) and Abbe and Montgomery (1996,

2003) on the Queets River.

2. Setting of the Quinault and Queets Rivers

2.1. Watershed physiography and climate

Both the Quinault River and Queets Rivers drain

the rugged core of the Olympic Mountains within

Olympic National Park before emptying into the

Pacific Ocean along the western coast of the Olympic

Peninsula (Fig. 1). The headwaters of each basin are

incised into Tertiary marine sedimentary and volcanic

rocks that have undergone rapid Cenozoic uplift

(Tabor and Cady, 1978a,b; Brandon et al., 1998).

Both rivers drain SW from glaciated mountains with

peaks higher than 2200 m, exiting the Olympic

Mountains and then flowing across a 10- to 30-km

wide coastal piedmont underlain by Quaternary gla-

ciofluvial sediment. The Queets River drains a total

area of about 1170 km2 and the Quinault River drains

1134 km2 (Table 1).

Fig. 1. Location maps of the study reaches. (a) Map of Queets and Quinault River drainage basins and geologic flood plains of the three study

reaches. Topographic base from USGS 30-m digital elevation model. (b) Geologic flood plains and 1994 channel of each of the three study

reaches. Flood-plain and channel transects shown at 1-km intervals. (c) Section of the Queets River flood plain showing 1994 channels and

flood-plain surfaces as mapped from 1994 orthophotos. Channel and flood-plain transects are spaced at 0.2-km intervals.

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–5932

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–59 33

Both of these watersheds are strongly influenced

by Pacific maritime conditions. During fall, winter,

and spring, these watersheds are repeatedly subject to

large storms from the SW, delivering substantial rain-

fall at lower elevations and snow in the higher

Olympic Mountains. Summers are relatively dry.

Mean annual precipitation is about 3.6 m at Lake

Quinault, somewhat less on the coast, and much

greater in the mountainous headwaters because of

orographic effects. The large precipitation volumes

are reflected in high average flows; average annual

flow generation is 3.4 m3/m2 for the Queets watershed

and about 3.7 m3/m2 for the Quinault watershed.

Particularly large flows, such as the one of March

1997, result during warm and wet storms during

which rivers gain substantial flow from melting of

low-elevation snowpack.

2.2. Alluvial reaches

As both rivers leave the core of the Olympic

Mountains and approach the western foothills and

coastal piedmont, valley bottoms broaden and the

rivers have established alluvial flood plains up to

3.5 km wide that are flanked by the western ramparts

of the Olympic Mountains or, further downstream, by

tall bluffs of Pleistocene glacial till and outwash.

Overall, channel gradient in the alluvial section of

the Queets River between river kilometer (RK) 0 and

RK 42.5 is 0.0023; the overall slope for the alluvial

section of Quinault River between RK 0 and RK 84.3

is 0.0022. Within their alluvial sections, both rivers

have cobble-gravel beds with pool-and-riffle profile

morphologies. The planview patterns of these rivers

do not neatly fit into standard classes such as ‘‘brai-

ded’’ or ‘‘meandering’’ but, at length scales of tens

of kilometers, are similar to the ‘‘irregular wander-

ing’’ or ‘‘irregular meandering’’ morphologies as des-

cribed by Church (1992). Shorter reaches within

these wandering river segments have braided (flow

around bars within an active channel) and anasto-

mosing (flow around islands excised from the flood

plain) planforms as defined by Knighton and Nanson

(1993).

Table 1

General study reach characteristics

Reach Lower Quinault Upper Quinault Queets

Location Lake Quinault outlet

(RK 51.6) to

Pacific Ocean

North Fork Quinault

River confluence

(RK 73.9) to

Lake Quinault

(RK 56.6)

Sams River

confluence

(RK 39.4) to

Pacific Ocean

Length (km)a 54.0 18.0 41.2

Drainage area (km2) 1124 at

downstream endb;

683 at upstream endc

441 at

upstream endb1152 at

River Mile 4.6c

Mean discharge (m3/s) 81c – 123c

Slope (m/m)d 0.0011 0.0035 0.0022

Low-flow channel width (m)a 65.4–27.7 58.3–24.3 68.9–39.6

Number of channelsa 1.02–0.14 1.41–0.73 1.30–0.62

Active channel width (m)a 95–45 240–104 165–89

Sinuosity (m/m)a 1.37 1.24 1.27

Mean flood plain width (m)e 1245 1930 1286

Total flood plain area (ha) 4831.8 2728.4 3914.3

Total area historically occupied

by channel (ha)f1509.5 675.2 1050.3

a Measured from 1994 orthophotoquads.b Geographical Information System analysis (Quinault Indian Nation and USDA Forest Service, 1999).c Information from USGS gage Quinault River at Quinault Lake (12039500), 1912–1997 and Queets River near Clearwater (12040500),

1931–1997 (Wiggins et al., 1998).d Measured from USGS 7.5-min topographic quadrangles.e Measured at 0.20-km increments along flood-plain centerline.f Cumulative area of channels depicted on historic maps and photos (Fig. 2).

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–5934

Within their alluvial valleys, both of these rivers

have shifted back and forth during the Holocene,

leaving a suite of fluvial landforms. Valley-bottom

surfaces range from unvegetated gravel bars to

densely forested alluvial surfaces several meters

above the present channel. Aquatic environments

include swift flowing main channels, side channels,

abandoned channels that are now ponds or wetlands,

and abandoned channels that now serve as routes for

local tributaries. The native flood-plain trees include

hardwoods such as red alder (Alnus rubra), vine

maple (Acer circinatum), bigleaf maple (Acer macro-

phyllum), and black cottonwood (Populus tricho-

carpa) and conifers Sitka spruce (Picea sitchensis),

western hemlock (Tsuga heterophylla), and smaller

amounts of western red cedar (Thuja plicata) and

Douglas fir (Pseudotsuga menziesii). Immense trees

grow in the temperate maritime climate, with stem

dimensions of some Sitka spruce, Douglas fir, western

red cedar exceeding 4 m in basal diameter and 70 m in

height.

Despite similarities in the geology, physiography,

and channel characteristics between the two water-

sheds, significant differences result in contrasting

channel and flood-plain processes and morphology.

The foremost dissimilarity is the continuity of sedi-

ment and wood transport. The Queets River flows

uninterrupted from its headwaters to the Pacific

Ocean, allowing continuous passage of water, sedi-

ment, and large woody debris along its entire course.

In contrast, the alluvial section of the Quinault River

is interrupted between RK 51.6 and RK 56.6 by Lake

Quinault, a moraine-dammed lake with a surface area

of 15 km2 (Fig. 1). Lake Quinault forms an inter-

mediate base level for the Quinault River and traps all

coarse sediment (sand and coarser) and most large

woody debris. Consequently, the Quinault River has

two distinct alluvial segments; an upstream segment

between RK 56.6 and RK 84.3 that receives sediment

and wood from upstream and is aggrading behind the

base level set by Lake Quinault, and a segment

downstream from the lake (RK 0 to RK 51.6) that

receives neither sediment nor wood from upstream.

Furthermore, transient flow storage in the lake sub-

stantially attenuates peak flows of the Quinault River

downstream of Lake Quinault, reducing peak dis-

charges of the five largest flows between 1933 and

1960 by 38%, as determined by comparing normal-

ized discharges for a series of floods on the Queets

and Quinault Rivers (Quinault Indian Nation and US

Department of Agriculture Forest Service, 1999, pp.

2.4–6).

Ownership and land use history also differ between

the Queets River and Quinault River valleys. Most of

the Queets River flood plain is now within the

Olympic National Park, although early settlers locally

cleared and harvested parts of the flood plain prior to

acquisition by the Park Service in the 1940s. The

lowermost 15 km of the Queets River is within the

Quinault Indian Nation Reservation, and the flanking

flood plain was mostly unaffected by timber harvest

until the 1990s. Upstream of Lake Quinault, the

Quinault River flood plain is under a mixture of

ownerships, including Olympic National Park, Olym-

pic National Forest, and private landowners. Parts

within the National Park are largely undisturbed, but

other ownerships have cleared and harvested small

areas of the flood plain. The Quinault River flood-

plain downstream from Lake Quinault is completely

within the Quinault Indian Nation Reservation and has

been almost completely harvested one or more times

since the late 1920s.

2.3. The study reaches

These differences and the general physiography

conveniently lead to the three independently analyzed

river segments (Fig. 1; Table 1): (i) the lower

Quinault River downstream of Lake Quinault with

no upstream-derived sediment or wood and an aver-

age gradient of 0.0011, (ii) the alluvial section of

upper Quinault River between the North Fork con-

fluence and Lake Quinault that does receive wood

and coarse sediment from upstream and has an

average gradient of 0.0035, and (iii) the alluvial

section of Queets River between RK 41 and RK 0

that also receives sediment and wood from upstream,

but has a lower average gradient of 0.0022. The

multiple differences among the reaches in what might

be considered ‘‘independent’’ or ‘‘predictor’’ varia-

bles (sediment and wood supply, slope, flood dis-

charge, land use history) prevent rigorous isolation of

the effects of individual variables on channel and

flood-plain morphology. Nevertheless, measured and

observed differences in channel pattern and dynam-

ics, as discussed in the following sections, can be

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–59 35

Table 2

Summary of sources for mapping of historic channel positions

Channel map source Scale Coverage Dates Plot

date

Comments

General Land Office

Surveys (available

at the Bureau of

Land Management,

Portland, OR)

1:31,680 Queets

Upper

Quinault

Lower

Quinault

Sams River

confluence to

Pacific Ocean

Entire study reach

Quinault Indian

Reservation

1895–1906

1895–1906

1902

1900

1900

1902

Instrument survey of

stream channel (and

unvegetated gravel bars?).

Digitized from 1:31,680

paper copies; georeferenced

by section corners.

Office of Indian Affairs

(available at Quinault

Indian Nation,

Taholah, WA)

1:31,680 Lower

Quinault

Quinault Indian

Reservation

Map dated 1920,

notes state

surveyed

1915–1917

1915 Transit and stadia for

primary control. Digitized

from 1:31,680 paper copy;

georeferenced by section

corners.

US Army Corps of

Engineer Tactical

Maps ‘‘Destruction

Island’’ and ‘‘Queets’’

1:62,500 Queets Elk Park to

Pacific Ocean

Maps dated 1922,

unknown

survey date

1922 Surveyed and compiled by

US Coastal and Geodetic

Survey; methods unknown.

Map not used for channel

migration analysis because

of unsatisfactory registration.

US Geological Survey

(USGS) Plan and

Profile River Maps

(available at University

of Washington Map

Library, Seattle, WA)

1:31,680 Queets

Upper

Quinault

Lower

Quinault

Paul Creek

confluence to

Pacific Ocean

Rustler Creek

confluence to

Lake Quinault

Lake Quinault

to Pacific Ocean

Map dated 1935,

notes state

surveyed

1931–1933

Map dated 1930,

notes state

surveyed 1929

1931

1929

1929

Survey methods unknown

but probably by transit and

stadia. Digitized from paper

copy; georeferenced by

section corners. Lower

Quinault reach required

significant editing due to

poor registration.

1939 USGS aerial

photographs (available

at US Geological

Survey, Tacoma, WA)

1:60,000 Most of the Olympic Peninsula 7–39 and 8–39 1939 1939 channel transcribed

by eye onto 7.5Vquadranglesand digitized. Spatially

referenced by quadrangle

corners.

15VUSGS topographic

quadrangles

1:62,500 Queets Entire study

reach

Maps dated 1956;

on the basis of 1952

source photographs

1952 Digitized from paper copy.

Spatially referenced by

quadrangle corners.

Upper

Quinault

No 1950s

source coverage

east of

123jE45jN(RK 67.6)

Maps dated

1953–1955

on the basis

of 1953–1954

aerial photography

1953

Lower

Quinault

Entire study

reach

1953

Duncan and Steinbrenner

(1976?) Soil Survey

(available at Quinault

Indian Nation, Taholah,

WA)

1:31,680 Lower

Quinault

Quinault Indian

Reservation

Map dated 1976(?);

on the basis of 1972

aerial photographs

1972 ‘‘River’’ polygons from

existing Quinault Indian

Nation digital coverage.

Date of source photographs

provided by Tony Hartrich,

Quinault Indian Nation.

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–5936

qualitatively attributed to differences in flow, slope,

and sediment and wood inputs.

3. Methods

For each of the three study reaches, channel

characteristics and historic channel migration rates

were mapped and measured from historic and current

aerial photographs and maps. Inferences regarding

interactions between large woody debris, standing

forests, the channel, and flood plain were based on

historical records and reconnaissance field observa-

tions, supplemented by previous work on the Queets

River by Fetherston et al. (1995), Abbe (2000), and

Abbe and Montgomery (1996, 2003).

3.1. Modern profile and planform characteristics

Channel profiles were obtained from spot water-

surface elevations and 6 m (20 ft) contour crossings

on USGS 7.5-min topographic quadrangles that were

surveyed during the 1980s. Modern channel plan-

form characteristics were measured from digital

orthophotos made in the summer of 1994 that

covered all three river segments (Table 2). To

quantify planform features, measurement transects

were placed perpendicular to the centerline of the

Table 2 (continued )

Channel map source Scale Coverage Dates Plot

date

Comments

7.5VUSGS topographic

quadrangles

1:24,000 Queets

Upper

Quinault

Lower

Quinault

Entire study

reach

Entire study

reach

Entire study

reach

Map dated

1982–1990; on

the basis of 1980,

1981, and 1987

source photographs

Maps dated 1990;

on the basis of 1981,

1985, and 1987

aerial photography

Map dated

1982–1990, on

the basis of 1980,

1981, 1985, and

1987 source

photographs

1981

1981

1981

Digitized from 1:24,000

paper copy. Spatially

referenced by quadrangle

corners. Plot dates assigned

on the basis of comparisons

of mapped channel positions

with aerial photographs from

time period.

1994 USGS and

Washington

Department of

Natural Resources

(DNR) digital

orthophotoquads

Derived

from

1:12,000

to 1:40,000

aerial photos

Queets

Upper

Quinault

Lower

Quinault

Entire study

reach

Entire study

reach

Entire study

reach

DNR Sept. 22, 1994

USGS Sept. 16, 1994

USGS July 11

(major;) and

DNR Sept. 22

(minor), 1994

1994 Channel digitized from

1:24,000 paper copy

(Queets R.) and on screen

from 1:5000) georeferenced

digital orthophotoquadrangles

(Quinault R.). Queets R.

discharge Sept. 22: 25.4 m3/s.

Quinault R. discharge July 11:

35.7 m3/s, Quinault R.

discharge Sept. 16: 54.4 m3/s,

Quinault R. discharge

Sept. 22: 24.5 m3/s

1997 Quinault Indian

Nation digital

orthophotoquads

(available at

Quinault Indian

Nation, Taholah, WA)

Derived

from

1:12,000

aerial

photos

Lower

Quinault

(and lower

15.8 km of the

Queets River)

Quinault Indian

Reservation

DNR 1997 Channel digitized on screen

(1:5000) from georeferenced

digital orthophotoquads. Only

the lower Quinault coverage

used in the channel migration

analysis.

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–59 37

primary low-flow channel at 0.2-km increments (Fig.

1c). The primary low-flow channel was defined as

the widest wetted channel, although in some loca-

tions of two or more roughly equal-sized channels,

we arbitrarily chose a single channel from which to

establish transect locations and orientations. For each

of these transects, we measured the number, width,

and area of the primary low-flow channel as well as

other visible water-filled channels connected at their

upstream and downstream termini at the time of the

photographs (dates and approximate discharges indi-

cated in Table 2). We also measured the number,

width, and area of isolated or partly isolated water-

bodies assumed not be flowing, defined as those

either not connected to a through-going low-flow

channel (i.e., flood-plain lakes and ponds) or a

channel only connected at one end to a through-

going low-flow channel (i.e., backwater channel). In

addition, we also measured the number, width, and

area of unvegetated gravel bars. We considered the

sum of the widths of the flowing channels and

flanking unvegetated gravel bars to represent the

‘‘active channel’’ width as described by Osterkamp

and Hedman (1982), which for the Quinault River

and Queets Rivers is the part of the flood plain that

had sufficient flow during the few years prior to the

summer of 1994 to prevent substantial establishment

of woody vegetation. The resolution of the 1994

orthophotos is 2 m, but, because of canopy cover, it

is likely that many channels and gravel bars less

than 10 m wide and bordered by vegetated flood

plains were not included in the transect measure-

ments.

More general flood-plain properties were charac-

terized with similar types of measurements using the

centerline of the valley bottom (here termed ‘‘geologic

flood plain’’) as a frame of reference (Fig. 1c). A

flood-plain reference frame allows for systematic

measurements of broad-scale attributes such as chan-

nel sinuosity and flood-plain width. Furthermore,

measurements based on a flood-plain centerline frame

of reference are more appropriate for channel and

flood-plain features that are widely dispersed across

the flood plain and not necessarily causally associated

with the present channel.

For both rivers, the geologic flood plain was map-

ped from aerial photographs, topographic maps, and

field observations and consists of the relatively flat

areas between flanking valley slopes. For the Queets

River, the geologic flood plain closely corresponds to

the Holocene alluvium map unit of Thackray (1996).

The mapped flood plains of both rivers contain areas of

slightly higher elevation that may be due to Neoglacial

or earlier Holocene aggradation but are nevertheless

inundated during large flows. The delineated flood

plains for all three river segments exclude tributary fans.

3.2. Channel migration

Patterns and rates of channel change for each of the

three study reaches were determined from maps, aerial

photos, and orthophotos showing channel position

between 1895 and 1997 (Table 2, Fig. 2). Channel

boundaries as shown on each source were digitized

into a geographic information system. Channel posi-

tions derived from maps and orthophotos were geore-

ferenced by quadrangle corners and public land survey

township corners. Channel positions from aerial photo-

graph sources were first transcribed onto topographic

quadrangles and then digitized.

Several sources of uncertainty affect quantitative

assessments of channel position with time from these

types of historical data. A key uncertainty associated

with the turn-of-the-century General Land Office

(GLO) surveys is that they do not necessarily show

the actual channel boundary but the inferred channel

extent at ‘‘mean high-water elevation,’’ which ‘‘is

found at the margin of the area occupied by the water

for the greater portion of each average year’’ (Bureau of

Land Management, 1973, pp. 93–97). In practice, this

definition probably results in local inclusion of gravel

bars and overflow channels outside the low-flow chan-

nel, whereas most other sources portray only the low-

flow channel. This was clearly the case for the GLO

surveys of the upper Quinault River (Fig. 2b), but the

GLO maps for the lower Quinault River and Queets

River distinguish between channel and unvegetated

gravel bars. Other uncertainties arise from differences

in flow stage, errors in transcription, and errors in

registration and digitizing. From consideration of the

degree of overlay of relatively stable reaches and the

scale of the source documents, our estimate of the

maximum error in placement of channel boundaries

from the older sources (1900–1902 GLOmaps, 1921–

1931 USGS Plan and Profile Maps, and 1939 aerial

photos) is about 50 m, but the error is probably much

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–5938

Fig. 2. Maps of historic channel positions as digitized from sources described in Table 2. Geologic flood plain indicated by unshaded area. For

the upper Quinault River study reach, there is no 1950 coverage east of longitude 123j45V.

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–59 39

less ( < 20 m) for the more recent topographic maps and

orthophotos. In the worst case, an uncertainty of 50 m

over a 10-years time interval between channel position

sources results in errors of about 50% to 100% of the

mean distance of channel movement for the three

reaches. However, for longer time intervals and higher

accuracy source maps, the errors are probably much

smaller.

For this study, we quantified channel migration

across the flood plain in two manners. The first

approach is similar to a variety of methods for

obtaining lineal measurements of lateral channel

movement (e.g., Leopold, 1973; Hickin and Nanson,

1975; Hooke, 1980; Pizzuto, 1994; Gurnell et al.,

1994; Gillespie and Giardino, 1996; Gurnell, 1997;

Elliot and Gyetvai, 1999; Shields et al., 2000). For

each of the three study reaches, we measured changes

in the position of the intersection of channel centerline

of the primary low-flow channel with flood-plain

transects spaced at 0.2-km increments, thus providing

70 (upper Quinault), 150 (Queets), and 196 (lower

Quinault) closely spaced measurements of lateral

channel movement (orthogonal to the flood-plain

axis) for each photo and map interval. This method

is similar to that of Gurnell et al. (1994), Piegay et al.

(1996), and Gurnell (1997), and allows for systematic

temporal and spatial analysis of channel movement

within a flood-plain frame of reference without the

bias typically introduced by trying to make measure-

ments at specific channel geometric features, such as

channel bends or straight reaches.

The second approach to evaluating channel move-

ment was to define each mapped channel and the

geologic flood plain as polygons in a geographic in-

formation system and evaluate the temporal sequence

of channel positions. This approach has been the ba-

sis of studies such as Piegay et al. (1996), Jacobson and

Pugh (1997), and Ham and Church (2000). For each

record of channel position, we calculated (i) the total

area of the low-flow channel, (ii) the area of channel

outside the area of the previous record, and (iii) the area

of channel outside any previous map or photo record of

previous channel positions. These results allow assess-

ment of how total channel area has changed with time

and the proportion of new channel area at each map

date that represents erosion of historically uneroded

flood plain (in contrast to reoccupying channel loca-

tions shown in previous maps and photos). Thus

providing a basis for determining flood-plain turnover

rates—that is, a measure of time required for the

channel to occupy the entire flood plain.

3.3. Large woody debris and forested flood-plain

interactions

The role of in-channel large woody debris and

flood-plain vegetation in affecting channel migration

was evaluated from historical accounts of locations

and positions of large wood jams, including records of

early settlers, notes accompanying late 1800s and

early 1900s General Land Office Surveys, and maps

and aerial photographs showing locations of wood

jams. Reconnaissance field mapping of channel con-

ditions, wood accumulations, bank stratigraphy and

vegetation conditions supplemented these historical

sources. Included in these field observations were

site visits before and after the flood of March 19,

1997 (1315 m3/s, Quinault River; 3115 m3/s Queets

River), which was the largest (Quinault River) or

second largest (Queets River) flood since 1955. Addi-

tionally, we relied heavily on the findings of Fether-

ston et al. (1995), Abbe and Montgomery (1996,

2003), and Abbe (2000) regarding the role of channel

and flood-plain wood in affecting channel morphology

and migration processes on the Queets River.

4. Results

4.1. Modern profile and planform characteristics

Measurements of 1994 flood-plain and channel

planform characteristics (Figs. 2 and 3) show distinct

differences among the three river segments, many of

Fig. 3. Box plots summarizing major study reach characteristics as measured from the channel and flood-plain transects. Boxes show 75th, 50th,

and 25th percentiles; vertical lines depict the 90th and 10th percentiles; individual symbols are points outside the 10th and 90th percentiles.

Low-flow channel width, active channel width, and number of low-flow channels are from measurements from 1994 orthophotos at 0.2-km

increments along the 1994 channel axis. Channel migration rate is from analysis of historic lateral movement of the primary channel measured

on transects orthogonal to the flood-plain axis spaced at 0.2-km increments.

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–5940

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–59 41

which that would be difficult to describe using typical

channel classification schemes (e.g., Leopold and

Wolman, 1957; Kellerhals et al., 1976; Schumm,

1985; Church, 1992; Knighton and Nanson, 1993).

The average flood-plain width for the lower Quinault

and Queets segments is about 1250 m, but the upper

Quinault River flood plain is substantially wider with

a mean width of 2470 m. Channel sinuosity is similar

for the upper Quinault River (1.24) and Queets River

(1.27) segments but is higher for the lower Quinault

River (1.37). All of these values are within the < 1.5

‘‘sinuous’’ (as opposed to >1.5 ‘‘meandering’’) cat-

egory of Leopold et al. (1964, p. 281).

Although resolution is coarse, all three reaches

have longitudinal profiles of generally decreasing

gradient downstream (Fig. 4a). Channel slopes for

both the upper Quinault and Queets segments

decrease almost monotonically, resulting in smooth

profiles at this scale of analysis. In contrast, the profile

of the lower Quinault River is more varied; gradients

range between 0.0001 and 0.0025, with less system-

atic change in the downstream direction.

Fig. 4. Plots of channel and flood-plain features for each of the three study reaches. The thin vertical lines relate the flood-plain axis used for

plots d– f to the 1994 channel axis used for plots b and c, and the 1980s channel axis used for plot a. Wide patterned vertical lines on the lower

Quinault River plots show relative locations of the three channel-spanning log jams shown on the 1915 timber survey map of the Quinault

Indian Reservation (Table 2). (a) Channel profile and reach gradients, from information on USGS 7.5Vtopographic quadrangles. (b) Low-flowand active channel width, as measured from 1994 orthophotos (Table 1) at 0.2-km increments along the axis of the primary low-flow channel.

Low-flow channels were defined as water-filled channels visibly connected at both their upstream and downstream ends on the 1994

orthophotos. The ‘‘active channel’’ was operationally defined to include the low-flow channels and flanking unvegetated sand and gravel bars.

(c) The number of low-flow channels and unvegetated gravel bars intersected by transects orthogonal to the 1994 low-flow channel at 0.2-km

increments. (d) Width of the Holocene flood plain as portrayed in Fig. 1, measured at 0.2-km increments along the flood-plain centerline. (e)

Number of low-flow and isolated channels, measured at 0.2-km increments along the flood-plain centerline. Isolated channels included all

flood-plain waterbodies visible on the 1994 orthophotos that were not low-flow channels. (f) Mean annual channel migration rate for the entire

period of historical channel information for each flood-plain transect.

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–5942

The low-flow channel widths of each of the three

river segments are similar, but the active channel

widths are significantly different (Figs. 3 and 4b).

The upper Quinault River has by far the widest active

channel, averaging 240 m wide compared to 165 m

for the Queets River and 95 m for the lower Quinault.

The active channel of upper Quinault River is almost

everywhere at least twice as wide as the low-flow

channel. However, for both for the Queets and lower

Quinault segments, several reaches more than a kilo-

meter long consist of a single channel flowing entirely

within vegetated flood-plain surfaces, with no flank-

ing exposed and unvegetated gravel bars. For both the

Queets and lower Quinault segments, the low-flow

channels widen as they approach within 1 to 2 km of

the Pacific Ocean.

Although all three river segments have reaches

containing multiple channels at low flow, the upper

Quinault and Queets segments have more low-flow

channels than the lower Quinault River, averaging

1.41 and 1.30 channels per flood-plain transect,

respectively, compared to 1.02 for the lower Quinault

River (Table 1; Figs. 3 and 4c,e). For both the upper

Quinault River and Queets River, there are several

reaches as long as a kilometer with two or three low-

flow channels. Some reaches have four or five visible

low-flow channels. The lower Quinault River has only

five multichannel reaches, all of which are shorter

than 400 m.

4.2. Patterns of channel movement, channel migra-

tion rates, and flood-plain turnover

The historic map and photo sequences show that

the styles and spatial patterns of channel migration

vary between the three study reaches (Fig. 2). The

channels of the upper Quinault River and the Queets

River upstream of RK 20 are mobile, with no stable

reaches and most migration accomplished by lateral

bank erosion and short avulsions (generally < 0.5 km)

involving low-amplitude meander curves with wave-

lengths of 2 to 3 km (Figs. 2 and 5). Downstream of

RK 20 on the Queets River, most of the historic

channel migration has been associated with progres-

sive enlargement and subsequent cutoff of five large

meander loops. Three of these meander loops have

migrated over more than a kilometer of flood-plain

width during the last 100 years and have undergone

Fig. 5. Sequence of channel maps of a portion of the Queets River near RK 26. Maps from 1895–1994 from sources listed in Table 2. The 1999

map from a field sketch.

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–59 43

complete cycles of growth and cutoff (Fig. 6). Rea-

ches between individual meander loops have re-

mained generally stable.

The lower Quinault River has a planview channel

pattern similar to the downstream portion of the

Queets River study reach, flowing through meanders

with wavelengths of up to 4 km and amplitudes as

great as 2 km (Fig. 1b). In contrast with the Queets

River, few of these meander loops have migrated

substantially during the last 100 years (Fig. 2a). Most

channel migration on the lower Quinault River has

been associated with (i) channel avulsions across

previously stable meander loops and (ii) adjacent

short stretches of more fully dynamic reaches similar

to the upper Quinault and upper Queets Rivers. Three

reaches have been particularly active historically: (i) a

short reach within RK 6–8; (ii) between RK 10 and

17; and (iii) a long reach between RK 27 and 40,

where, since 1902, the channel has avulsed across two

meander bends (Fig. 7).

Channel migration rates vary within and among the

three study segments (Figs. 3 and 4f). Calculated on

the basis of changes in channel centerline position

relative to flood-plain transects spaced at 0.2-km

increments, the mean channel migration rate for the

lower Quinault River between 1902 and 1997 was

5.0F 3.9 m/year. The Queets River had a similarly

calculated mean migration rate of 5.6F 4.5 m/year

between 1900 and 1994. The upper Quinault River

has had a channel migration rate of 8.8F 4.1 m/year

between 1902 and 1994, significantly higher than

those measured for the lower Quinault and Queets

segments ( pb0.01; two-tailed t-test). Error terms are

the standard deviation of the spatial variation in mean

migration rate, as averaged for each flood-plain trans-

ect for the entire period of record.

Owing to substantial back-and-forth channel

movement between times of known channel position,

the calculated migration rates will likely be less than

the actual migration rates by some factor that depends

on the length of time between sequential maps. Con-

sequently, the calculated migration rates for each of

the study reaches is likely biased by the different

number of historic sources and uneven interval

lengths between known channel positions. For each

of the three study segments, significant inverse corre-

lations ( p < 0.1; two-tailed t-test) between measured

reach-average migration rates and the number of years

between map dates (Fig. 8) indicate that this bias is

indeed present for all three reaches and that a signifi-

cant amount of channel migration is ‘‘missed’’

between times of known channel positions. To

account for this, we have normalized migration rates

to a 17-year interval period, representing the mean

interval between all pairs of channel position infor-

mation used in the study, using the linear regressions

Fig. 6. Sequence of channel maps of a portion of the Queets River near RK 14. Sources listed in Table 2.

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–5944

shown in Fig. 8. This adjustment results in normalized

migration rates of 4.0F 1.2 m/year for the lower

Quinault River, 7.5F 2.9 m/year for the Queets River,

and 12.7F 3.3 m/year for the upper Quinault River

(the error terms are the estimated standard error for a

sampling interval of 17 years resulting from the

regression of the mean cross-section migration rate

against the sampling period, calculated using Eq.

1.4.7 of Draper and Smith, 1966).

The other approach to analyzing channel change,

the overlay of channel polygons, shows that for all

three reaches, the apparent channel areas were greatest

at the time of the turn-of-the-century General Land

Office mapping (Fig. 9). These earliest maps, however,

undoubtedly include large active channel areas outside

the low-flow channels—especially for the upper Qui-

nault River and perhaps for the Queets River. For the

1939 and later sources, which were all based on aerial

photographs made during summer low-flow periods,

channel area for each of the three study reaches has

varied by less than F 20%. For almost all time periods

on the upper Quinault and Queets segments, more than

50% of the channel area at a specific map date was

outside of the channel area of the previous record. For

the upper Quinault River, channel movement between

times of mapping typically involved more than 75% of

the channel area (Fig. 9b). For the lower Quinault

River, the area of new channel between map dates

has generally been less than 50% of the total channel

area (Fig. 9a). Similarly, the proportion of channel

movement that was into previously unoccupied flood

plain (as opposed to previous channel locations) has

remained relatively constant for the last 50 to 60 years.

(The results from the first part of the historical record

are skewed by the shortness of the record.) For the

Queets and upper Quinault segments, 40% to 50% of

the channel erosion between photo and map sets is into

historically unoccupied flood plain; whereas for the

lower Quinault River, only 18% to 30% of new channel

areas are outside of historical channel positions.

Over the 94 to 97 years covered by the channel

maps, the channel has at some time occupied 25% to

Fig. 7. Sequence of channel maps of a portion of the lower Quinault River at the site of the 3rd log jam near RK 38. Sources listed in Table 2.

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–59 45

29% of the geologic flood plains (Fig. 2). The annual

rate with which the channels moved into flood-plain

areas not occupied previously within the record of

maps and photos has generally varied between about

0.05% and 0.5% per year (Fig. 10). Extrapolated

forward, these rates imply 200- to 2000-year flood-

plain ‘‘turnover’’ periods in which, on average, the

channel occupies all of the geologic flood plain.

However, mobile channels typically form and

erode relatively young flood-plain surfaces near to

recent and present channel positions, allowing surfa-

ces farther away to become much older (Everitt,

1968). This is the case for both the Queets River

and Quinault River, where 18% to 50% of channel

migration reworks flood plain or gravel bars < 100

years old (Fig. 9). Therefore, the age distribution of

flood-plain surfaces is likely to be logarithmic, lead-

ing Everitt (1968) and Gottesfeld and Gottesfeld

(1990) to discuss the time scale of flood-plain turn-

over in terms of flood-plain half-lives, defined as the

length of time in which the channel occupies half of

the total flood-plain area.

We have calculated flood-plain half-lives for each

of the three study reaches by fitting exponential

decay curves to the cumulative erosion of 1900–

1902 flood plains (areas outside the 1900–1902

mapped channel, but within the geologic flood plain)

over the subsequent 94 to 97 years (Fig. 11). For the

Queets River segment, the calculated flood-plain

half-life is 385 + 48/� 39 years (the stated error

represents the standard error about the exponential

regression), whereas the half-life is somewhat longer

at 495 + 38/� 32 years for the upper Quinault River

flood plain and somewhat shorter at 277 + 53/� 38

years for the lower Quinault River segment,

although this latter result is heavily influenced by

the measured channel differences between the 1902

and 1915 maps for which errors due to poor

registration may be large. Because of the likelihood

that there were areas of the flood plains occupied by

the channels over the last 94 to 97 years that were

not recorded at the isolated instances portrayed on

the channel maps, these calculated half-lives likely

overestimate the actual flood-plain half-lives, al-

Fig. 8. Scatter plots and linear regressions showing relations between the interval between successive channel maps used in the analysis and the

calculated annual channel migration rate for the corresponding periods. Vertical gray line represents the mean interval (17 years) between

successive channel maps used in this analysis and is the basis for marmalizing the calculated channel migration rates.

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–5946

though this error is probably partly compensated for

by registration errors associated with the individual

maps and photos (which would most likely lead to

erroneously high calculated migration rates). This

approach to calculating flood-plain half-life differs

from the more direct approach adopted by Everitt

(1968) in which the age distribution of the flood-

plain surface was determined by mapping tree ages.

Such an approach could be undertaken on the

Queets and Quinault Rivers to test the results

determined from analysis of channel movement, as

well as provide information on flood-plain turnover

rates prior to human land use effects, but this has

not yet been done.

Fig. 9. Results of GIS analysis of sequential channel positions for the (a) lower Quinault, (b) upper Quinault, and (c) Queets River segments

showing, for each map/photo record, (i) the area of channel, (ii) the area of channel that is different than the previous record of channel

positions, and (iii) the area of channel that is different than any previous map/photo record.

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–59 47

4.3. Large woody debris

On the lower Quinault River, despite little wood

entering from upstream, large in-channel log jams

were common. Early settlers reported that wood accu-

mulations grew large enough to completely span the

channel, including jams that were nearly 0.5 km long

and completely impassable to navigation (Cleland,

1959, pp. 177–178). Prior to the historic record, such

jams required the Quinault Indians to build flat-bot-

tomed canoes to ease portage over these jams when

navigating the Quinault River (Capoeman, 1991, p.

Fig. 11. Exponential decay curves fit to the cumulative erosion (as indicated by subsequent channel positions shown in Fig. 2) of the portions of

the geologic flood plains that were outside the channels portrayed on the 1900–1902 General Land Office maps.

Fig. 10. Temporal variation in mean annual erosion of the geologic flood plain by the channel. Calculated on the basis of annual percent area of

channel movement into areas of the flood plain, which have not been historically occupied by the channel. Results are partly a function of the

length of record, especially for the first 40 to 50 years (when the historical record is relatively short).

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–5948

Fig. 12. Typical settings of in-channel wood, sediment, and flood-plain vegetation on the three river segments. (a) April 2, 1997, view

downstream from the right bank of the upper Quinault River at RK 67.6, showing single pieces and small groups of wood that have accumulated

on gravel bars between low-flow channels (analogous to the ‘‘bar-top’’ jams of Abbe and Montgomery, 1996). Channel is flanked by stands of

red alder. (b) Aug. 7, 1999, view upstream of the Queets River at RK 26. Location and orientation of the view shown in Fig. 5. This log jam and

associated aggradation of the former channel (off to the right in this perspective) resulted in a 700-m avulsion of the channel during the March

1997 flood along a route marked by a small overflow channel on 1994 aerial photos (Fig. 5). (c) April 29, 1997, view downstream of the lower

Quinault River from left bank at RK 16.6. This has been a reach of historic channel migration near the ‘‘2nd log jam’’ depicted on the 1915–

1916 timber surveys (Table 2). The March 1997 flood resulted in formation of a point bar of gravel and wood, forcing flow to erode laterally

into the adjacent flood plain. Flood-plain vegetation, consisting primarily of even-aged stands of red alder and groves of spruce has been

entrained and deposited on the downstream bar.

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–59 49

107). A map constructed during a 1915–1916 timber

survey (Table 2) of the reservation shows three large

log jams (Fig. 2a), one of which covered about 15 ha

(Fig. 7). Presently, no channel log jams span the lower

Quinault River, in part due to active removal to aid

navigation (Philip Martin Jr., Quinault Indian Nation,

oral communication, 1997) and, perhaps, less large

wood in the river because of substantial flood-plain

timber harvest during the last 80 years.

The sequence of events at what was mapped as

‘‘3rd log jam’’ on the 1915–1916 timber survey (Fig.

7) may be representative of channel and flood-plain

conditions associated with channel-spanning log jams

on the lower Quinault River prior to significant human

channel clearing and flood-plain timber harvest. The

location of this jam is consistent with early 1890s

accounts of a ‘‘big jam of trees and rubbish com-

pletely blocking the channel’’ (Cleland, 1959, pp.

177–178). This large jam completely spanned the

channel in 1902, according to notes accompanying

the General Land Office survey of the channel (cadas-

tral survey notes for T. 22 N., R. 10 E., housed at the

Bureau of Land Management office in Portland,

Oregon). The 1902 General Land Office survey

depicts a wide and ponded channel at the area of the

jam but no note or portrayal of a bypass channel. The

1915–1916 timber survey also shows the log jam

completely blocking the main channel but by that

time, a small bypass channel had formed. A 1929

survey of the channel shows that the bypass channel

had become the main channel, with the former route

dashed in, indicating as much as 500 m of lateral

avulsion of nearly a kilometer of channel. A similar

sequence of events apparently caused another avul-

sion in the next meander downstream between 1953

and 1997 (Fig. 7), partly resulting from a log jam most

likely formed of wood eroded from the flood-plain

forest during the 1902–1929 avulsion. The locations

of other two log jams portrayed on the 1915–1916

timber survey have also had similar but smaller

avulsions and have been reaches of persistent channel

instability (Figs. 2 and 4).

In contrast to the large, channel-spanning jams and

consequent channel avulsions that have affected at

least three reaches of the lower Quinault River during

the early 20th century, the Queets River and upper

Quinault River (and, to a more local extent, the lower

Quinault River) have mostly been affected by indi-

vidual pieces of wood and wood jams that flank or

only partially block the channel (Fig. 12). These wood

accumulations have been associated with rapid bar

growth, bank erosion, and short-length channel avul-

sions as described for the Queets River by Fetherston

et al. (1995), Abbe and Montgomery (1996, 2003),

and Abbe (2000). In these reaches of frequent wood

and sediment transport, large woody debris is com-

monly deposited on bar tops, at bar apices, and in

meander bends, commonly promoting mid-channel

and point bar growth which then redirects flow against

channel margins, causing bank erosion and channel

migration (Abbe and Montgomery, 1996). An exam-

ple on the Queets River is shown in Figs. 5 and 12b,

where about 500 m of new channel formed as a result

of partial channel blockage by wood and sediment,

resulting in a meander cutoff during the high flow of

March 1997. Two similar avulsions, both involving

channel lengths less than 400 m, occurred during the

same flood on the lower Quinault River at RK 32.5

and 27.2. Judging from the abundant in-channel

wood, our field observations of channel change

between 1996 and 1999, and high lateral migration

rates, such channel shifting and lateral migration is

common, perhaps occurring most years on the upper

Quinault River and Queets River above RK 20.

Downstream of RK 20 on the Queets River, the

channel has a lower gradient and more sinuous plan-

form and there has been more systematic growth and

cutoff of large meander loops (e.g., Fig. 6). Inspection

of aerial photographs and observations from site visits

indicate that these meanders grow as a result of point

Fig. 13. Photographs illustrating evidence for role of log jams in promoting flood-plain conifer growth. (a) Sitka spruce trees growing from

partly decomposed nurse logs at the site of the 1902–1915 3rd log jam on the lower Quinault River near RK 38. Shovel (0.5 m) is leaning

against a spruce stem. (b) Conifer seedlings growing on spruce stem deposited near the margin of the 1952 channel of the Queets River near RK

26. The largest is a 5-m tall spruce growing from the root wad of the nurse log. (c) Recently deposited spruce bole on the right margin of the

lower Quinault River at about RK 26. A 1.5-m tall spruce is growing from the root wad. (d) Grove of mature spruce growing at western margin

of the 1902–1915 3rd log jam on the lower Quinault River near RK 38. Shovel (0.5 m) is leaning against a spruce stem. (e) Exposure of the

right bank of the Quinault River near RK 21.8 where three Sitka spruce, including the 1.5-m diameter spruce on the left edge of the photograph,

apparently grew from a partially decomposed conifer log now emerging from the eroding bank (upon which the person is standing).

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–5950

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–59 51

bars forming on the insides of meander bends,

facilitated by deposition in the lee of wood accumu-

lations. The resulting aggradation, as indicated by

buried wood accumulations and wide active channels

within the meander bends, probably facilitates over-

bank flow across the meander bend and eventual

cutoff.

Longer-term feedbacks between fluvially trans-

ported wood and flood-plain vegetation are evident

at sites of historic wood accumulation and channel

change. At the location of the channel-spanning 3rd

log jam on the lower Quinault River segment, which

is now as far as 500 m away from the channel,

numerous spruce are now growing from higher logs

within this jam (Fig. 13a). Many of these spruce have

diameters of 30 to 40 cm within vegetation otherwise

dominated by maple and alder, indicating that these

log jams may facilitate flood-plain conifer establish-

ment by providing elevated and nourishing sites for

seedling establishment. Consistent with this, large

conifers recently deposited along both the Quinault

River and Queets River have young spruce and west-

ern red cedar growing from their stems and rootwads

(Fig. 13a,b). In addition, the unharvested flood plain

within the cutoff meander loop adjacent to the area of

the 3rd log jam is vegetated primarily by hardwoods,

but locally interspersed are groves of large spruce,

some with basal diameters exceeding 2 m (Fig. 13d).

These spruce grow in clusters of about 400 m2 or less

on surfaces that stand 1 to 2 m above the surrounding

flood plain. No direct evidence of nurse logs was

found at these groves, but commonly three or more of

the spruce would be aligned, supporting the inference

that these spruce groves have formed on long-rotted or

buried nurse logs. The grouping of these spruce, in

combination with the higher terrain which they are

found, leads us to conclude that these groves of

mature spruce occupy sites of old log jams, although

it cannot be ruled out that windthrow may have

formed the downed logs upon which the present

spruce have grown. A bank exposure downstream

along the lower Quinault River shows that similar

arrangements of large spruce are indeed rooted in logs

now buried by overbank sediment and forest soils

(Fig. 13e).

Similar interactions between channel shifting, wood

deposition and recruitment, and flood-plain morphol-

ogy and vegetation are also evident in conjunction with

wood accumulations that do not completely span the

channel. In particular for the Queets River, Fetherston

et al. (1995) described feedbacks between riparian

forest conditions and large woody debris accumula-

tions, and Abbe and Montgomery (1996) conducted

detailed studies between RK 41 and 46where they have

related wood accumulations to hydraulic, depositional,

and aquatic habitat conditions. These earlier studies on

the Queets River and our observations on all three river

segments indicate that deposited large woody debris

commonly promotes mid-channel and point bar growth

(Fig. 12a). Additionally, bar formation in the lee of

wood accumulations can lead to establishment of

riparian vegetation on the resulting higher and finer-

grained deposits, and ultimately, stands of even-age

riparian vegetation. Bank erosion and the consequent

entrainment of flood-plain trees can foster further

channel instability as the downed wood steers flow

into banks (the ‘‘auto diversion’’ process of Keller and

Swanson, 1979).

5. Discussion

Despite the overall similar physiographic settings

of the Queets River and Quinault River basins, the

three study reaches have distinct planform character-

istics as well as patterns and rates of channel migra-

tion and flood-plain erosion. These differences can be

logically attributed to the different sediment, water,

and wood inputs to the basin, and interactions among

these factors, although rigorous identification of cause

and effect is hampered by the multiple differences

between the reaches and the long time scales of

interactions. These inferences regarding effects of

sediment, wood, and flow on channel morphology,

coupled with observations of flood-plain morphology

and vegetation that may represent even longer time

frames, serve as a basis for speculation on interactions

among channel migration, in-channel wood, and

flood-plain forests at time scales ranging from decades

to centuries (Fig. 14).

5.1. Flood-plain and channel planform character-

istics

Variation in flood-plain width primarily reflects

Quaternary geologic controls on valley morphology.

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–5952

For both the lower Quinault River and Queets River

segments, there is pronounced narrowing of the flood

plains where they pass through Quaternary terminal

moraine complexes (Thackray, 1996). Within the

upper Quinault River segment, the flood-plain nar-

rows where bedrock ridges encroach from the north,

but widens where it is aggrading behind moraine-

dammed Lake Quinault. Flood-plain width does not

correlate with channel planform characteristics such

as low-flow width, active channel width, or the

frequency of low-flow channels ( p>0.1 in all cases;

two-tailed t-test). Thus, we infer that most aspects of

channel and flood-plain morphology are independent

of the local geologic controls that affect flood-plain

width.

Channel planform characteristics appear to be more

directly related to overall sediment, wood, and flow

regimes. For example, similar low-flow widths for all

three study reaches are likely due to similar low-flow

discharges. The greater active channel width, greater

channel slopes, higher migration rates, and greater

frequency of multiple and unconnected channels of

the upper Quinault River, and to a lesser extent the

Queets River, probably owe to the greater sediment

supply and higher peak discharges of these two rea-

ches compared to the lower Quinault River. The lower

Quinault River has an overall narrower active channel

width, lower channel gradient, and greater sinuosity,

likely reflecting the combined influence of upstream

sediment and wood trapping and flow attenuation

provided by Lake Quinault. These differences are

consistent with Schumm’s (1985) qualitative observa-

tions of the relationships among channel slope, sedi-

ment load, channel stability, and channel form.

Different base-level conditions may also be an

important factor in controlling planform channel and

flood-plain morphology. The Queets and upper Qui-

nault Rivers have overall similar sediment and flow

regimes; however, the upper Quinault River, which

has been aggrading behind Lake Quinault during the

Holocene, has a significantly wider flood plain, wider

active channel, more low-flow and high-flow chan-

nels, and greater migration rates. Valley aggradation

of the upper Quinault River has apparently primarily

affected the downstream 5 km of valley bottom, re-

sulting in lower channel gradient and higher channel

migration rate compared to the upstream part of the

upper Quinault study segment (Fig. 4).

In contrast, the Queets River and lower Quinault

Rivers have similar base-level conditions, but differ-

ent sediment, wood, and peak discharge conditions

because of the presence of Lake Quinault upstream of

Fig. 14. Feedbacks hypothesized between large wood and flood-plain and channel evolution, at time scales of decades to centuries, for the

Queets River and Quinault River. Partly derived from findings described by Fetherston et al. (1995), Abbe and Montgomery (1996) and Abbe

(2000).

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–59 53

the lower Quinault River. The relatively greater sedi-

ment and wood supply of the Queets River in con-

junction with greater peak flows is probably

responsible for its greater active channel width and

perhaps gradient as well. The greater sediment supply

and steeper channel of the Queets River also likely

result in the slightly greater channel migration rates

and, in conjunction with the wider active channel,

more numerous low-flow channels (which generally

occupy recently abandoned channels). In general,

though, the overall morphologic differences between

the Queets and lower Quinault segments are less

pronounced than are the differences between the

aggrading upper Quinault River and the other two

study reaches (Fig. 3).

5.2. Channel migration patterns and dynamics

Parts of all three study reaches have distinctive

styles and rates of channel migration (Figs. 2–4).

The upper Quinault River has been multi-channeled

with high migration rates, with most migration accom-

plished by lateral bank erosion and short avulsions.

Similarly, the channel of Queets River upstream of RK

20 has been everywhere active, but changes down-

stream to a series of stationary meander loops that have

grown and cutoff over time scales of about 100 years.

Channel migration on the lower Quinault River during

the last 94 years has been focused in three distinct

reaches of channel avulsions across meander loops and

adjacent reaches of lateral channel migration and short

avulsions. Between these dynamic reaches, the lower

Quinault River has been mostly stable.

On all three river segments, rates of channel migra-

tion over the last 100 years correspond with modern

channel and flood-plain planform characteristics (Fig.

4). This is especially evident for the Queets River and

lower Quinault River, where discrete reaches of dis-

tinctly higher historic migration rates have wider

active channel widths, greater numbers of low-flow

and unconnected channels, and more unvegetated

gravel bars compared to reaches that have been histor-

ically stable or less active. This correspondence sup-

ports the importance of channel migration as a process

responsible for many ecologically important channel

and flood-plain characteristics.

Despite different channel migration rates, the

flood-plain occupancy rates for all three reaches are

broadly similar, indicating that flood-plain turnover

occurs over 200 to 2000 years with flood-plain half-

lives of 200 to 500 years. These half-lives are close

to those determined for a variety of other alluvial

rivers (e.g., Nanson and Beach, 1977; Hughes, 1997),

but somewhat longer than the 30–50 years estimated

by Gottesfeld and Gottesfeld (1990) for the Morice

River in British Columbia, which has a similar

physiographic setting as the three study segments

analyzed here. Three factors apparently control

flood-plain turnover rates the Queets and Quinault

study segments: (i) the width of the flood plain; (ii)

the channel migration rate; (iii) the propensity of

channel migration to erode into the flood plain rather

than to occupy previous channel locations. These

factors apparently counterbalance each other in the

Queets and Quinault Rivers to produce broadly

similar flood-plain turnover rates. The high migration

rate of the upper Quinault River is offset by its wider

flood plain. In contrast, the Queets and lower Qui-

nault segments have lower migration rates but nar-

rower flood plains.

5.3. Interactions among wood, sediment, flow and

resulting channel and flood-plain morphology

At length scales of individual channel bends and

time scales of individual flow events to decades, many

of the channel migration and flood-plain growth and

erosion processes on all study reaches are controlled

by interactions between flow, sediment, and wood

(Fig. 12). As described for the Queets River by

Fetherston et al. (1995) and Abbe and Montgomery

(1996, 2003), and Abbe (2000), these processes

involve many forms of wood, including standing

trees, entrained wood, and previously deposited accu-

mulations of large woody debris. In places, wood

deposition triggers sedimentation, such as in the ‘‘bar

apex jams,’’ and in other locations, sediment deposits

trap woody debris, such as the ‘‘bar top jams’’ (Abbe

and Montgomery, 1996). In both cases, the resulting

accumulations of wood and sediment can promote

flood-plain generation by providing sites and substrate

suitable for establishment of flood-plain vegetation

(Gottesfeld and Gottesfeld, 1990; Abbe and Mont-

gomery, 1996; Fetherston et al., 1995). Additionally,

sediment and wood accumulations locally cause

flood-plain erosion by directing flow toward banks,

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–5954

thus introducing additional sediment and wood in the

river (e.g., Keller and Swanson, 1979). These inter-

actions are evident along most of the lengths of the

Queets and upper Quinault Rivers, which both receive

substantial sediment and wood from upstream as well

as from bank erosion, although the style of migration

apparently corresponds to channel slope on the Queets

River. All of these and similar processes have been

described for rivers with forested flood plains (e.g.,

Hack and Goodlet, 1960; Nanson and Beach, 1977;

Keller and Swanson, 1979; Swanson and Lienkaem-

per, 1982; Sedell and Froggatt, 1984; Gottesfeld and

Gottesfeld, 1990; Church, 1992; Piegay, 1993; Maser

and Sedell, 1994, pp. 44–50; Piegay and Gurnell,

1997; Piegay and Marston, 1998; Jacobson et al.,

1999; Gurnell et al., 2000).

On the lower Quinault River, however, these types

of short reach- and time-scale interactions have been

primarily restricted to the parts of the river that have

been historically active between RK 6 and 8, RK 10

and 17, and RK 27 and 40 (Fig. 2). This condition of

dynamic reaches within otherwise more stable chan-

nels has been observed in other rivers with forested

flood plains where they have been called ‘‘sedimen-

tation zones’’ (Church, 1983) or ‘‘disturbance rea-

ches’’ (Jacobson and Pugh, 1997). Higher channel

mobility within such sedimentation zones has been

attributed to external controls such as Neoglacial

sediment production, tributary fan locations (Church,

1983), and valley geometry (Jacobson and Pugh,

1997), as well as internal mechanisms involving

feedbacks between large woody debris accumula-

tions, sedimentation, and lateral channel shifting

(Keller and Swanson, 1979; Gottesfeld and Gottes-

feld, 1990). On the lower Quinault River, where

external controls such as tributary fans and upper

basin sediment and wood production have little influ-

ence on the channel or flood plain, internal mecha-

nisms involving wood, sediment, and channel

migrations must be the primary factors controlling

channel and flood-plain dynamics.

The correspondence of the three large log jams

mapped in the 1915–1916 timber survey with these

dynamic reaches on the lower Quinault River points to

the likely role of large log jams in creating persistent

zones of unstable channels (Fig. 4). In particular, the

long reach of high channel migration since 1902

between RK 27 and 40 is mostly downstream of the

large channel-spanning log jam at the time of the 1902

cadastral survey (Fig. 7). Ponding behind this log jam

likely triggered the channel avulsion between 1915

and 1929, and indirectly caused the 1953–1981 avul-

sion downstream (Fig. 7) as erosion of the new channel

brought additional wood from the flood-plain forest

into the channel, leading to formation of another jam

downstream. Wood and sediment brought into the

channel by these two jam-related avulsions has prob-

ably been the source of the continued downstream

instability in a river system where the only sources of

wood and sediment are the flood plains and valley

margins. For the lower Quinault River, such feedbacks

between flood-plain sediment and wood erosion and

deposition must be primary mechanisms promoting

channel instability.

The absence of large, channel-spanning log jams

on the upper Quinault River or Queets River may be

due to their greater sediment supply, peak discharge,

and gradient. Although there is no direct evidence for

this, we speculate that the higher sediment flux on the

upper Quinault and Queets River segments results in

more rapid sediment deposition at sites of wood

accumulation compared to the lower Quinault River,

thus forcing flow to erode laterally prior to complete

blockage of the channel by wood. The greater slopes

and peak discharges of the upper Quinault River and

Queets River would also promote such lateral erosion

as well as erosion of alternate flow routes prior to

complete blockage. In contrast, the lower Quinault

River, with low sediment flux and lower peak dis-

charges, may allow substantial wood accumulation—

to the point of completely blocking the channel—

before the ponded river completely overtops the flood

plain and erodes another course.

Channel-spanning log jams and consequent avul-

sions similar to those of the lower Quinault River have

been reported for other large, low-gradient rivers with

forested flood plains. Examples include the Willamette

River of western Oregon, where historic accounts

describe extensive rafts of wood blocking channels

and consequent formation of new channels (Sedell and

Froggatt, 1984); the Morice River of British Columbia

(Gottesfeld and Gottesfeld, 1990); the Skagit and

Stillaquamish Rivers of western Washington (Maser

and Sedell, 1994, p. 134); and the Red River of

Louisiana, where a complex series of channel-span-

ning log jams formed between the mid-17th century

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–59 55

and 1873, obstructing the channel for more than 250

km and diverting the lower Red River to another route

to the Gulf of Mexico (Veatch, 1906).

Feedbacks among log jams, channel migration,

and flood-plain vegetation also involve longer time

scales (e.g., Naiman et al., 2000). Bars composed of

relatively fine-grained sediment form in the lee of

large woody debris accumulations (Fig. 12a; Swan-

son and Lienkaemper, 1982; Gottesfeld and Gottes-

feld, 1990; Church, 1992; Maser and Sedell, 1994,

pp. 47–50; Fetherston et al., 1995; Abbe and Mont-

gomery, 1996, 2003; Jacobson et al., 1999). Woody

debris accumulations, especially large and stable bar-

apex and meander jams (Fig. 12b; Abbe and Mont-

gomery, 1996, 2003), and their associated down-

stream pendant bars commonly stand several meters

above adjacent channel and bar areas and provide

advantageous substrate and topographic conditions

for flood-plain vegetation establishment and growth.

For the Queets and Quinault Rivers, red alder typi-

cally colonizes these bars within a few years of

emplacement and, if the log jam remains in place,

will survive to further stabilize the bar and promote

soil and flood-plain forest development, resulting in

forest patches older than surrounding vegetation

growing in subsequently abandoned channels (Fether-

ston et al., 1995; Abbe and Montgomery, 1996;

Abbe, 2000).

Furthermore, the log jams themselves may become

important in controlling future in-channel wood and

flood-plain forest conditions. Log jams are a rich

source of organic material and provide habitats and

energy sources for a variety of terrestrial plant and

animal species (Maser and Sedell, 1994, pp. 47–50;

Fetherston et al., 1995; Jacobson et al., 1999). More-

over, log jams become germination sites for conifers,

especially spruce that grow directly on logs or from

soil clinging to rootwads on less-traveled tree stems

incorporated into the log jam (Fig. 13; Gottesfeld and

Gottesfeld, 1990). In the Queets and Quinault River

valley bottoms, it appears that large log jams provide

germination sites, stability, and organic-rich substrate

to allow conifers to successfully compete and coexist

with rapid hardwood colonization of the more barren

gravel bars (Fig. 13). This is shown by the young

spruce growing from members of the 1902 log jam on

the Quinault River, and perhaps by the groves of large

and aligned spruce at the margins of the 1902 jam.

Analogously, spruce forests advance into salt marshes

in coastal environments by colonizing stable piles of

driftwood (Maser and Sedell, 1994, p. 84).

We further speculate that for the Quinault River

and Queets River, such spruce groves may have

effects on channel migration on time scales of cen-

turies to millennia (Fig. 14) by (i) providing a source

of large trees that may eventually fall into the channel

and become key members of future log jams (Abbe

and Montgomery, 1996; Abbe, 2000) and (ii) forming

patches of flood-plain vegetation resistant to erosion,

thus affecting locations of future log jams and channel

migration. This may have been the case for the 1902

log jam on the lower Quinault River, for which the

downstream logs are racked up against one of these

mature spruce groves.

5.4. Management implications

These results and inferences have implications for

management of large, gravel-bed rivers flowing

through forested flood plains. Foremost, flood-plain

morphology and channel conditions at any one time

are the product of several hundred to thousands of

years of channel migration and flood-plain forest

development. Consequently, time scales of this order

are the necessary perspective from which to consider

the full suite of channel and flood-plain processes for

these types of rivers. River management strategies that

do not fully consider the entire flood plain over time

scales of centuries are likely to lead to eventual

changes in channel and flood-plain processes and

morphology (Naiman et al., 2000), although such

effects may require several decades or centuries to

become evident. In contrast, most present strategies

for actively managed forested flood plains implicitly

operate in consideration of time frames of at most

several decades, corresponding to timber cutting

schedules and areas of sediment and wood input

associated with the present channel location. Sec-

ondly, channel and flood-plain processes and mor-

phologies can be very different, even within broadly

similar physiographic settings such as the western

Olympic Peninsula. For example, the distribution

and extent of various flood-plain environments (chan-

nels, unvegetated gravel bars, flood-plain forests of

various composition and age structure) were likely to

have been different for each of the three study reaches

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–5956

under natural conditions because of differences in the

natural sediment, flow, and wood regimes. Specifica-

tion of realistic goals and optimizing actions to meet

habitat or harvest objectives must account for these

differences and the processes that underlie them.

6. Conclusions

The gravel-bed channels and forested flood plains

of the lower Quinault River, upper Quinault River,

and Queets River have distinctive physical morphol-

ogies and dynamics that reflect different flow, sedi-

ment, and wood regimes within an overall similar

physiographic environment. These are the major

observations:

(i) Local channel migration rates, active channel

width, and number of channels are highly

correlated to one another (Fig. 4). The sedi-

ment-and-wood-rich upper Quinault River, ag-

grading behind Lake Quinault, has the greatest

migration rate, widest active channel, most multi-

channel reaches, and greatest overall channel

slope of the three river segments. The lower

Quinault River, with no upstream sediment

source, has the lowest migration rate, narrowest

active channel, fewest multi-channel reaches, and

the lowest overall gradient. The Queets River,

with a sediment supply that is likely similar to the

upper Quinault River but in a nonaggradational

setting, has attributes intermediate between the

upper Quinault and lower Quinault segments.

(ii) Overall, rates of encroachment of the channel

across the flood plain are broadly similar for each

of the three study reaches, resulting in flood-plain

half-lives of 200 to 500 years. The higher channel

migration rate of the upper Quinault River

relative to the Queets and lower Quinault Rivers

is compensated by a wider flood plain. For the

lower Quinault River segment, the lower overall

channel migration rate and the relatively low

proportion of channel movement into historically

unoccupied flood plain are likely responsible for

the overall narrower Holocene flood plain.

(iii) Historic patterns and processes of channel

migration are different among the three reaches,

and these differences influence channel and

flood-plain morphology as well as spatial

variations in channel migration and flood-plain

occupancy rates. Channel change on the upper

Quinault River and the upstream portion of the

Queets River study reach is primarily due to

short-wavelength, low-amplitude meander for-

mation and cutoffs, largely spurred by bar

formation in the lee of large woody debris

accumulations and by partial blockage of the

channel by log jams. These reaches are dynamic

and multichanneled, and channels commonly

shift back and forth between previously occu-

pied channels on an annual to decadal time

frame. For the lower portion of the Queets River

study reach, most channel migration is due to

progressive enlargement and cutoff of large

meander bends, with complete cycles of mean-

der growth and cutoff occurring over several

decades to centuries. For the lower Quinault

River, avulsions are apparently triggered by

channel-spanning log jams, which then set up

reaches of channel instability separated by long

reaches of relatively stable channels.

(iv) Channel migration, sediment transport, large

woody debris, and flood-plain vegetation and

morphology involve several feedbacks, some

encompassing several century durations of the

same order as flood-plain turnover rates. Channel

shifting introduces large woody debris and flood-

plain sediment into the channel system, which

then triggers further channel shifting. For the

lower Quinault River, this is probably the primary

process maintaining input of wood and sediment

into the channel. Log jams and sediment accu-

mulations become unique terrestrial environ-

ments for flood-plain forest formation. The log

jams are preferred sites of conifer germination and

thus become areas resistant to future flood-plain

erosion as well as areas that may eventually

provide source wood for future log jams.

Acknowledgements

Phillip Martin Jr. of the Quinault Indian Nation,

Department of Natural Resources, provided logistical

support and boat access to the lower Quinault River.

Mark Mobbs, Justine Turner, Antonio Hartrich, and

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–59 57

John Sims of the Quinault Indian Nation provided

historic documents and maps. Tina Marie and Danial

Polette, US Geological Survey, provided cartographic

and GIS assistance. Discussions with Mark Mobbs,

Dave Lasorda, Gordon Grant, Tim Abbe, Jonathan

Friedman, and Fred Swanson helped to clarify many

of the ideas presented here. Reviews of earlier

versions by John Parker, Ed Prych, Jonathan Fried-

man, and Fred Swanson improved the paper. Kevin

Fetherston and an anonymous referee provided help-

ful reviews for Geomorphology, as did journal editors

David Montgomery and Herve Piegay. Partial funding

was provided by the Quinault Indian Nation.

References

Abbe, T.B., 2000. Patterns, mechanics and geomorphic effects of

wood debris accumulations in a forest river system. PhD dis-

sertation. University of Washington, Seattle.

Abbe, T.B., Montgomery, D.R., 1996. Large woody debris jams,

channel hydraulics and habitat formation in large rivers. Regu-

lated Rivers: Research and Management 12, 201–221.

Abbe, T.B., Montgomery, D.R., 2003. Patterns and processes of

wood debris accumulation in the Queets River Basin, Washing-

ton. Geomorphology, this issue.

Brandon, M.T., Roden-Tice, M.K., Garver, J.I., 1998. Late Ceno-

zoic exhumation of the Cascadia accretionary wedge in the

Olympic Mountains, northwest Washington State. Geological

Society of America Bulletin 110, 985–1009.

Bureau of Land Management, 1973. Manual of Instructions for the

Survey of the Public Lands of the United States. Bureau of Land

Management Technical Bulletin 6, Catalog Number I 53.7:973-

2. US Government Printing Office, Washington, DC.

Capoeman, P.K. (Ed.), 1991. Land of the Quinault, 2nd ed. Qui-

nault Indian Nation, Taholah, WA.

Church, M., 1983. Pattern of instability in a wandering gravel bed

channel. Special Publication of the International Association of

Sedimentologists 6, 169–180.

Church, M., 1992. Channel morphology and typology. In: Carlow,

P., Petts, G.E. (Eds.), The Rivers Handbook, vol. 1. Blackwell,

Oxford, UK, pp. 126–143.

Cleland, L.H., 1959. Trials and Trails of the Pioneers of the Olym-

pic Peninsula, State of Washington. Humptulips Pioneers Asso-

ciation, Hoquiam, WA.

Draper, N.R., Smith, H., 1966. Applied Regression Analysis. Wiley,

New York.

Duncan, S.H., Steinbrenner, E.C., 1976. Soil Survey, Quinault In-

dian Reservation. Quinault Indian Nation, Taholah, WA. 12

plates (1:31,600). 13 pp.

Elliot, J.G., Gyetvai, S., 1999. Channel-Pattern Adjustments and

Geomorphic Characteristics of Elkhead Creek, Colorado,

1937–1997. U.S. Geological Survey Water-Resources Investi-

gations, Report 99-4098. US Government Printing Office, Den-

ver, CO.

Everitt, B.L., 1968. Use of the cottonwood in an investigation of the

recent history of a flood plain. American Journal of Science 266,

417–439.

Fetherston, K.L., Naiman, R.J., Bilby, R.E., 1995. Large woody

debris, physical process, and riparian forest development in

montane river networks of the Pacific Northwest. Geomorphol-

ogy 13, 133–144.

Gillespie, B.M., Giardino, J.R., 1996. Determining the migratory

activity index for a river: an example from the Brazos River,

Texas. Zeitschrift fur Geomorphologie 40, 417–428.

Gottesfeld, A.S., Gottesfeld, L.M.J., 1990. Floodplain dynamics of

a wandering river, dendrochronology of the Morice River, Brit-

ish Columbia, Canada. Geomorphology 3, 159–179.

Gurnell, A.M., 1997. Channel change on the River Dee meanders,

1946–1992, from the analysis of air photographs. Regulated

Rivers: Research and Management 13, 13–26.

Gurnell, A.M., Downward, S.R., Jones, R., 1994. Channel planform

change on the River Dee meanders, 1976–1992. Regulated

Rivers: Research and Management 9, 187–204.

Gurnell, A.M., Petts, G.E., Hannah, D.M., Smith, B.P.G., Edwards,

P.J., Kollmann, J., Ward, J.V., Tockner, K., 2000. Wood storage

within the active zone of a large European gravel-bed river.

Geomorphology 34, 55–72.

Hack, J.T., Goodlet, J.C., 1960. Geomorphology and Forest Ecol-

ogy of a Mountain Region in the Central Appalachians. U.S.

Geological Survey Professional Paper, vol. 347. US Govern-

ment Printing Office, Washington, DC.

Ham, D.G., Church, M., 2000. Bed-material transport estimated

from channel morphodynamics, Chilliwack River, British Co-

lumbia. Earth Surface Processes and Landforms 25, 1123–1142.

Hickin, E.J., 1984. Vegetation and river channel dynamics. Cana-

dian Geographer 28, 111–126.

Hickin, E.J., Nanson, G.C., 1975. The character of channel migra-

tion on the Beatton River, Northeast British Columbia, Canada.

Geological Society of America Bulletin 86, 487–494.

Hooke, J.M., 1980. Magnitude and distribution of rates of river

bank erosion. Earth Surface Processes 5, 143–157.

Hughes, F.M.R., 1997. Floodplain biogeomorphology. Progress in

Physical Geography 21, 501–529.

Jacobson, R.B., Pugh, A.L., 1997. Riparian-Vegetation Controls

on the Spatial Pattern of Stream-Channel Instability, Little

Pine Creek, Missouri. U.S. Geological Survey Water-Supply

Paper, vol. 2494. US Government Printing Office, Washing-

ton, DC.

Jacobson, P.J., Jacobson, K.M., Angermeier, P.L., Cherry, D.S.,

1999. Transport, retention, and ecological significance of woody

debris within a large ephemeral river. Journal of the North

American Benthological Society 18, 429–444.

Keller, E.A., Swanson, F.J., 1979. Effects of large organic material

on channel form and fluvial processes. Earth Surface Processes

4, 361–380.

Kellerhals, R., Church, M., Bray, D.I., 1976. Classification and

analysis of river processes. Journal of the Hydraulics Division

102, 813–829.

Knighton, A.D., Nanson, G.C., 1993. Anastomosis and the contin-

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–5958

uum of channel pattern. Earth Surface Processes and Landforms

18, 613–625.

Leopold, L.B., 1973. River channel change with time—an example.

Geological Society of America Bulletin 84, 1845–1860.

Leopold, L.B., Wolman, M.G., 1957. River Channel Patterns—

Braided, Meandering and Straight. United States Geological

Survey Professional Paper, vol. 282B. US Government Printing

Office, Washington, DC, pp. 39–85.

Leopold, L.B., Wolman, M.G., Miller, J.P., 1964. Fluvial Processes

in Geomorphology. Freeman, San Francisco, CA.

Maser, C., Sedell, J.R., 1994. From the Forest to the Sea—the

Ecology of Wood in Streams, Rivers, Estuaries and Oceans.

St. Lucie Press, Delray Beach, FL.

Naiman, R.J., Elliott, S.R., Helfield, J.M., O’Keefe, T.C., 2000.

Biophysical interactions and the structure and dynamics of riv-

erine ecosystems: the importance of biotic feedbacks. Hydro-

biologia 410, 79–86.

Nanson, G.C., Beach, H.F., 1977. Forest succession and sedimen-

tation on a meandering-river floodplain, northeast British Co-

lumbia, Canada. Journal of Biogeography 4, 229–251.

Osterkamp, W.R., Hedman, E.R., 1982. Perennial-Streamflow Char-

acteristics Related to Channel Geometry and Sediment in Mis-

souri River Basin. U.S. Geological Survey Professional Paper,

vol. 1242. US Government Printing Office, Washington, DC.

Piegay, H., 1993. Nature, mass and preferential sites of coarse woody

debris deposit in the Lower Ain valley (Mollon reach) France.

Regulated Rivers: Research and Management 8, 359–372.

Piegay, H., Gurnell, A.M., 1997. Large woody debris and river

geomorphological pattern: examples from S.E. France and S.

England. Geomorphology 19, 99–116.

Piegay, H., Marston, R.A., 1998. Distribution of large woody debris

along the outer bend of meanders in the Ain River, France.

Physical Geography 19, 318–340.

Piegay, H., Barge, O., Landon, N., 1996. Streamway Concept

Applied to River Mobility/Human Use Conflict Management.

Proceedings ‘‘Rivertech96’’ 1st International Conference on

New/Emerging Concepts for Rivers, Chicago, USA, Septem-

ber, pp. 681–688.

Pizzuto, J.E., 1994. Channel adjustments to changing discharges,

Powder River, Montana. Geological Society of America Bulle-

tin 106, 1494–1501.

Quinault Indian Nation and US Department of Agriculture Forest

Service, 1999. Quinault River Watershed Analysis. Quinault

Indian Nation, Taholah, WA.

Schumm, S.A., 1985. Patterns of alluvial rivers. Annual Review of

Earth and Planetary Science 13, 5–27.

Sedell, J.R., Froggatt, J.L., 1984. Importance of streamside forests

to large rivers: the isolation of the Willamette River, Oregon,

USA, from its floodplain by snagging and streamside forest

removal. Internationale Vereinigung fur Theoretische and Ange-

wandte Limnologie Verhandlungen (International Association of

Theoretical and Applied Limnology) 22, 1828–1834.

Shields Jr., F.D., Simon, A., Steffen, L.J., 2000. Reservoir effects on

downstream river channel migration. Environmental Conserva-

tion 27, 54–66.

Swanson, F.J., Lienkaemper, G.W., 1982. Interactions among flu-

vial processes, forest vegetation, and aquatic ecosystems, South

Fork Hoh River, Olympic National Park. In: Starkey, E.E.,

Franklin, J.F., Matthew, J.W. (Eds.), Ecological Research in Na-

tional Parks of the Pacific Northwest. Proceedings of the 2nd

Conference on Scientific Research in the National Parks, No-

vember 1979, San Francisco, California. Oregon State Univer-

sity Forest Research Laboratory, Corvallis, OR, pp. 30–34.

Tabor, R.W., Cady, W.M., 1978a. The Structure of the Olympic

Mountains, Washington. Analysis of a Subduction Zone. U.S.

Geological Survey Professional Paper, vol. 1033. US Govern-

ment Printing Office, Washington, DC.

Tabor, R.W., Cady, W.M., 1978b. Geologic Map of the Olympic

Peninsula. U.S. Geological Survey Miscellaneous Investigations

Map I-994. US Government Printing Office, Washington, DC.

Thackray, G.D., 1996. Glaciation and neotectonic deformation on

the western Olympic Peninsula, Washington. PhD dissertation.

University of Washington, Seattle.

Veatch, A.C., 1906. Geology and Underground Water Resources of

Northern Louisiana and Southern Arkansas. U.S. Geological

Survey Professional Paper, vol. 46. U.S. Government Printing

Office, Washington, DC.

Wiggins, W.D., Ruppert, G.P., Smith, R.R., Reed, L.L., Courts,

M.L., 1998. Water Resources Data, Washington, Water Year

1997. US Geological Survey Water-Data Report WA-97-1, Ta-

coma, WA.

J.E. O’Connor et al. / Geomorphology 51 (2003) 31–59 59